static int cbq_enqueue(struct ifaltq *ifq, struct mbuf *m, struct altq_pktattr *pktattr) { cbq_state_t *cbqp = (cbq_state_t *)ifq->altq_disc; struct rm_class *cl; struct pf_mtag *t; int len; IFQ_LOCK_ASSERT(ifq); /* grab class set by classifier */ if ((m->m_flags & M_PKTHDR) == 0) { /* should not happen */ #if defined(__NetBSD__) || defined(__OpenBSD__)\ || (defined(__FreeBSD__) && __FreeBSD_version >= 501113) printf("altq: packet for %s does not have pkthdr\n", ifq->altq_ifp->if_xname); #else printf("altq: packet for %s%d does not have pkthdr\n", ifq->altq_ifp->if_name, ifq->altq_ifp->if_unit); #endif m_freem(m); return (ENOBUFS); } cl = NULL; if ((t = pf_find_mtag(m)) != NULL) cl = clh_to_clp(cbqp, t->qid); #ifdef ALTQ3_COMPAT else if ((ifq->altq_flags & ALTQF_CLASSIFY) && pktattr != NULL) cl = pktattr->pattr_class; #endif if (cl == NULL) { cl = cbqp->ifnp.default_; if (cl == NULL) { m_freem(m); return (ENOBUFS); } } #ifdef ALTQ3_COMPAT if (pktattr != NULL) cl->pktattr_ = pktattr; /* save proto hdr used by ECN */ else #endif cl->pktattr_ = NULL; len = m_pktlen(m); if (rmc_queue_packet(cl, m) != 0) { /* drop occurred. some mbuf was freed in rmc_queue_packet. */ PKTCNTR_ADD(&cl->stats_.drop_cnt, len); return (ENOBUFS); } /* successfully queued. */ ++cbqp->cbq_qlen; IFQ_INC_LEN(ifq); return (0); }
/* * priq_enqueue is an enqueue function to be registered to * (*altq_enqueue) in struct ifaltq. */ static int priq_enqueue(struct ifaltq *ifq, struct mbuf *m, struct altq_pktattr *pktattr) { struct priq_if *pif = (struct priq_if *)ifq->altq_disc; struct priq_class *cl; struct pf_mtag *t; int len; IFQ_LOCK_ASSERT(ifq); /* grab class set by classifier */ if ((m->m_flags & M_PKTHDR) == 0) { /* should not happen */ printf("altq: packet for %s does not have pkthdr\n", ifq->altq_ifp->if_xname); m_freem(m); return (ENOBUFS); } cl = NULL; if ((t = pf_find_mtag(m)) != NULL) cl = clh_to_clp(pif, t->qid); #ifdef ALTQ3_COMPAT else if ((ifq->altq_flags & ALTQF_CLASSIFY) && pktattr != NULL) cl = pktattr->pattr_class; #endif if (cl == NULL) { cl = pif->pif_default; if (cl == NULL) { m_freem(m); return (ENOBUFS); } } #ifdef ALTQ3_COMPAT if (pktattr != NULL) cl->cl_pktattr = pktattr; /* save proto hdr used by ECN */ else #endif cl->cl_pktattr = NULL; len = m_pktlen(m); if (priq_addq(cl, m) != 0) { /* drop occurred. mbuf was freed in priq_addq. */ PKTCNTR_ADD(&cl->cl_dropcnt, len); return (ENOBUFS); } IFQ_INC_LEN(ifq); /* successfully queued. */ return (0); }
struct mbuf * ip6_forward(struct mbuf *m, struct route_in6 *ip6forward_rt, int srcrt) { struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); struct sockaddr_in6 *dst; struct rtentry *rt; int error, type = 0, code = 0; boolean_t proxy = FALSE; struct mbuf *mcopy = NULL; struct ifnet *ifp, *origifp; /* maybe unnecessary */ u_int32_t inzone, outzone; struct in6_addr src_in6, dst_in6; #if IPSEC struct secpolicy *sp = NULL; #endif struct timeval timenow; unsigned int ifscope = IFSCOPE_NONE; #if PF struct pf_mtag *pf_mtag; #endif /* PF */ getmicrotime(&timenow); #if PF pf_mtag = pf_find_mtag(m); if (pf_mtag != NULL && pf_mtag->pftag_rtableid != IFSCOPE_NONE) ifscope = pf_mtag->pftag_rtableid; /* * If the caller provides a route which is on a different interface * than the one specified for scoped forwarding, discard the route * and do a lookup below. */ if (ifscope != IFSCOPE_NONE && (rt = ip6forward_rt->ro_rt) != NULL) { RT_LOCK(rt); if (rt->rt_ifp->if_index != ifscope) { RT_UNLOCK(rt); rtfree(rt); rt = ip6forward_rt->ro_rt = NULL; } else { RT_UNLOCK(rt); } } #endif /* PF */ #if IPSEC /* * Check AH/ESP integrity. */ /* * Don't increment ip6s_cantforward because this is the check * before forwarding packet actually. */ if (ipsec_bypass == 0) { if (ipsec6_in_reject(m, NULL)) { IPSEC_STAT_INCREMENT(ipsec6stat.in_polvio); m_freem(m); return (NULL); } } #endif /*IPSEC*/ /* * Do not forward packets to multicast destination (should be handled * by ip6_mforward(). * Do not forward packets with unspecified source. It was discussed * in July 2000, on ipngwg mailing list. */ if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 || IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { ip6stat.ip6s_cantforward++; /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */ if (ip6_log_time + ip6_log_interval < timenow.tv_sec) { ip6_log_time = timenow.tv_sec; log(LOG_DEBUG, "cannot forward " "from %s to %s nxt %d received on %s\n", ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst), ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif)); } m_freem(m); return (NULL); } if (ip6->ip6_hlim <= IPV6_HLIMDEC) { /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */ icmp6_error(m, ICMP6_TIME_EXCEEDED, ICMP6_TIME_EXCEED_TRANSIT, 0); return (NULL); } /* * See if the destination is a proxied address, and if so pretend * that it's for us. This is mostly to handle NUD probes against * the proxied addresses. We filter for ICMPv6 here and will let * icmp6_input handle the rest. */ if (!srcrt && nd6_prproxy) { VERIFY(!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)); proxy = nd6_prproxy_isours(m, ip6, ip6forward_rt, ifscope); /* * Don't update hop limit while proxying; RFC 4389 4.1. * Also skip IPsec forwarding path processing as this * packet is not to be forwarded. */ if (proxy) goto skip_ipsec; } ip6->ip6_hlim -= IPV6_HLIMDEC; /* * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU - * size of IPv6 + ICMPv6 headers) bytes of the packet in case * we need to generate an ICMP6 message to the src. * Thanks to M_EXT, in most cases copy will not occur. * * It is important to save it before IPsec processing as IPsec * processing may modify the mbuf. */ mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN)); #if IPSEC if (ipsec_bypass != 0) goto skip_ipsec; /* get a security policy for this packet */ sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, IP_FORWARDING, &error); if (sp == NULL) { IPSEC_STAT_INCREMENT(ipsec6stat.out_inval); ip6stat.ip6s_cantforward++; if (mcopy) { #if 0 /* XXX: what icmp ? */ #else m_freem(mcopy); #endif } m_freem(m); return (NULL); } error = 0; /* check policy */ switch (sp->policy) { case IPSEC_POLICY_DISCARD: case IPSEC_POLICY_GENERATE: /* * This packet is just discarded. */ IPSEC_STAT_INCREMENT(ipsec6stat.out_polvio); ip6stat.ip6s_cantforward++; key_freesp(sp, KEY_SADB_UNLOCKED); if (mcopy) { #if 0 /* XXX: what icmp ? */ #else m_freem(mcopy); #endif } m_freem(m); return (NULL); case IPSEC_POLICY_BYPASS: case IPSEC_POLICY_NONE: /* no need to do IPsec. */ key_freesp(sp, KEY_SADB_UNLOCKED); goto skip_ipsec; case IPSEC_POLICY_IPSEC: if (sp->req == NULL) { /* XXX should be panic ? */ printf("ip6_forward: No IPsec request specified.\n"); ip6stat.ip6s_cantforward++; key_freesp(sp, KEY_SADB_UNLOCKED); if (mcopy) { #if 0 /* XXX: what icmp ? */ #else m_freem(mcopy); #endif } m_freem(m); return (NULL); } /* do IPsec */ break; case IPSEC_POLICY_ENTRUST: default: /* should be panic ?? */ printf("ip6_forward: Invalid policy found. %d\n", sp->policy); key_freesp(sp, KEY_SADB_UNLOCKED); goto skip_ipsec; } { struct ipsec_output_state state; /* * All the extension headers will become inaccessible * (since they can be encrypted). * Don't panic, we need no more updates to extension headers * on inner IPv6 packet (since they are now encapsulated). * * IPv6 [ESP|AH] IPv6 [extension headers] payload */ bzero(&state, sizeof(state)); state.m = m; state.dst = NULL; /* update at ipsec6_output_tunnel() */ error = ipsec6_output_tunnel(&state, sp, 0); key_freesp(sp, KEY_SADB_UNLOCKED); if (state.tunneled == 4) return (NULL); /* packet is gone - sent over IPv4 */ m = state.m; if (state.ro.ro_rt) { rtfree(state.ro.ro_rt); state.ro.ro_rt = NULL; } if (error) { /* mbuf is already reclaimed in ipsec6_output_tunnel. */ switch (error) { case EHOSTUNREACH: case ENETUNREACH: case EMSGSIZE: case ENOBUFS: case ENOMEM: break; default: printf("ip6_output (ipsec): error code %d\n", error); /* fall through */ case ENOENT: /* don't show these error codes to the user */ break; } ip6stat.ip6s_cantforward++; if (mcopy) { #if 0 /* XXX: what icmp ? */ #else m_freem(mcopy); #endif } m_freem(m); return (NULL); } } skip_ipsec: #endif /* IPSEC */ dst = (struct sockaddr_in6 *)&ip6forward_rt->ro_dst; if ((rt = ip6forward_rt->ro_rt) != NULL) { RT_LOCK(rt); /* Take an extra ref for ourselves */ RT_ADDREF_LOCKED(rt); } if (!srcrt) { /* * ip6forward_rt->ro_dst.sin6_addr is equal to ip6->ip6_dst */ if (rt == NULL || !(rt->rt_flags & RTF_UP) || rt->generation_id != route_generation) { if (rt != NULL) { /* Release extra ref */ RT_REMREF_LOCKED(rt); RT_UNLOCK(rt); rtfree(rt); ip6forward_rt->ro_rt = NULL; } /* this probably fails but give it a try again */ rtalloc_scoped_ign((struct route *)ip6forward_rt, RTF_PRCLONING, ifscope); if ((rt = ip6forward_rt->ro_rt) != NULL) { RT_LOCK(rt); /* Take an extra ref for ourselves */ RT_ADDREF_LOCKED(rt); } } if (rt == NULL) { ip6stat.ip6s_noroute++; in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute); if (mcopy) icmp6_error(mcopy, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOROUTE, 0); m_freem(m); return (NULL); } RT_LOCK_ASSERT_HELD(rt); } else if (rt == NULL || !(rt->rt_flags & RTF_UP) || !IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &dst->sin6_addr) || rt->generation_id != route_generation) { if (rt != NULL) { /* Release extra ref */ RT_REMREF_LOCKED(rt); RT_UNLOCK(rt); rtfree(rt); ip6forward_rt->ro_rt = NULL; } bzero(dst, sizeof(*dst)); dst->sin6_len = sizeof(struct sockaddr_in6); dst->sin6_family = AF_INET6; dst->sin6_addr = ip6->ip6_dst; rtalloc_scoped_ign((struct route *)ip6forward_rt, RTF_PRCLONING, ifscope); if ((rt = ip6forward_rt->ro_rt) == NULL) { ip6stat.ip6s_noroute++; in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute); if (mcopy) icmp6_error(mcopy, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOROUTE, 0); m_freem(m); return (NULL); } RT_LOCK(rt); /* Take an extra ref for ourselves */ RT_ADDREF_LOCKED(rt); } /* * Source scope check: if a packet can't be delivered to its * destination for the reason that the destination is beyond the scope * of the source address, discard the packet and return an icmp6 * destination unreachable error with Code 2 (beyond scope of source * address) unless we are proxying (source address is link local * for NUDs.) We use a local copy of ip6_src, since in6_setscope() * will possibly modify its first argument. * [draft-ietf-ipngwg-icmp-v3-04.txt, Section 3.1] */ src_in6 = ip6->ip6_src; if (in6_setscope(&src_in6, rt->rt_ifp, &outzone)) { /* XXX: this should not happen */ ip6stat.ip6s_cantforward++; ip6stat.ip6s_badscope++; m_freem(m); return (NULL); } if (in6_setscope(&src_in6, m->m_pkthdr.rcvif, &inzone)) { ip6stat.ip6s_cantforward++; ip6stat.ip6s_badscope++; m_freem(m); return (NULL); } if (inzone != outzone && !proxy) { ip6stat.ip6s_cantforward++; ip6stat.ip6s_badscope++; in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard); if (ip6_log_time + ip6_log_interval < timenow.tv_sec) { ip6_log_time = timenow.tv_sec; log(LOG_DEBUG, "cannot forward " "src %s, dst %s, nxt %d, rcvif %s, outif %s\n", ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst), ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp)); } /* Release extra ref */ RT_REMREF_LOCKED(rt); RT_UNLOCK(rt); if (mcopy) { icmp6_error(mcopy, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_BEYONDSCOPE, 0); } m_freem(m); return (NULL); } /* * Destination scope check: if a packet is going to break the scope * zone of packet's destination address, discard it. This case should * usually be prevented by appropriately-configured routing table, but * we need an explicit check because we may mistakenly forward the * packet to a different zone by (e.g.) a default route. */ dst_in6 = ip6->ip6_dst; if (in6_setscope(&dst_in6, m->m_pkthdr.rcvif, &inzone) != 0 || in6_setscope(&dst_in6, rt->rt_ifp, &outzone) != 0 || inzone != outzone) { ip6stat.ip6s_cantforward++; ip6stat.ip6s_badscope++; m_freem(m); return (NULL); } if (m->m_pkthdr.len > rt->rt_ifp->if_mtu) { in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig); if (mcopy) { uint32_t mtu; #if IPSEC struct secpolicy *sp2; int ipsecerror; size_t ipsechdrsiz; #endif mtu = rt->rt_ifp->if_mtu; #if IPSEC /* * When we do IPsec tunnel ingress, we need to play * with the link value (decrement IPsec header size * from mtu value). The code is much simpler than v4 * case, as we have the outgoing interface for * encapsulated packet as "rt->rt_ifp". */ sp2 = ipsec6_getpolicybyaddr(mcopy, IPSEC_DIR_OUTBOUND, IP_FORWARDING, &ipsecerror); if (sp2) { ipsechdrsiz = ipsec6_hdrsiz(mcopy, IPSEC_DIR_OUTBOUND, NULL); if (ipsechdrsiz < mtu) mtu -= ipsechdrsiz; key_freesp(sp2, KEY_SADB_UNLOCKED); } /* * if mtu becomes less than minimum MTU, * tell minimum MTU (and I'll need to fragment it). */ if (mtu < IPV6_MMTU) mtu = IPV6_MMTU; #endif /* Release extra ref */ RT_REMREF_LOCKED(rt); RT_UNLOCK(rt); icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu); } else { /* Release extra ref */ RT_REMREF_LOCKED(rt); RT_UNLOCK(rt); } m_freem(m); return (NULL); } if (rt->rt_flags & RTF_GATEWAY) dst = (struct sockaddr_in6 *)(void *)rt->rt_gateway; /* * If we are to forward the packet using the same interface * as one we got the packet from, perhaps we should send a redirect * to sender to shortcut a hop. * Only send redirect if source is sending directly to us, * and if packet was not source routed (or has any options). * Also, don't send redirect if forwarding using a route * modified by a redirect. */ if (!proxy && ip6_sendredirects && rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) { if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) != 0) { /* * If the incoming interface is equal to the outgoing * one, and the link attached to the interface is * point-to-point, then it will be highly probable * that a routing loop occurs. Thus, we immediately * drop the packet and send an ICMPv6 error message. * * type/code is based on suggestion by Rich Draves. * not sure if it is the best pick. */ RT_REMREF_LOCKED(rt); /* Release extra ref */ RT_UNLOCK(rt); icmp6_error(mcopy, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADDR, 0); m_freem(m); return (NULL); } type = ND_REDIRECT; } #if IPFW2 /* * Check with the firewall... */ if (ip6_fw_enable && ip6_fw_chk_ptr) { u_short port = 0; ifp = rt->rt_ifp; /* Drop the lock but retain the extra ref */ RT_UNLOCK(rt); /* If ipfw says divert, we have to just drop packet */ if (ip6_fw_chk_ptr(&ip6, ifp, &port, &m)) { m_freem(m); goto freecopy; } if (!m) { goto freecopy; } /* We still have the extra ref on rt */ RT_LOCK(rt); } #endif /* * Fake scoped addresses. Note that even link-local source or * destinaion can appear, if the originating node just sends the * packet to us (without address resolution for the destination). * Since both icmp6_error and icmp6_redirect_output fill the embedded * link identifiers, we can do this stuff after making a copy for * returning an error. */ if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) { /* * See corresponding comments in ip6_output. * XXX: but is it possible that ip6_forward() sends a packet * to a loopback interface? I don't think so, and thus * I bark here. ([email protected]) * XXX: it is common to route invalid packets to loopback. * also, the codepath will be visited on use of ::1 in * rthdr. (itojun) */ #if 1 if (0) #else if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0) #endif { printf("ip6_forward: outgoing interface is loopback. " "src %s, dst %s, nxt %d, rcvif %s, outif %s\n", ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst), ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp)); } /* we can just use rcvif in forwarding. */ origifp = m->m_pkthdr.rcvif; } else origifp = rt->rt_ifp; /* * clear embedded scope identifiers if necessary. * in6_clearscope will touch the addresses only when necessary. */ in6_clearscope(&ip6->ip6_src); in6_clearscope(&ip6->ip6_dst); ifp = rt->rt_ifp; /* Drop the lock but retain the extra ref */ RT_UNLOCK(rt); /* * If this is to be processed locally, let ip6_input have it. */ if (proxy) { VERIFY(m->m_pkthdr.aux_flags & MAUXF_PROXY_DST); /* Release extra ref */ RT_REMREF(rt); if (mcopy != NULL) m_freem(mcopy); return (m); } #if PF /* Invoke outbound packet filter */ error = pf_af_hook(ifp, NULL, &m, AF_INET6, FALSE, NULL); if (error != 0 || m == NULL) { if (m != NULL) { panic("%s: unexpected packet %p\n", __func__, m); /* NOTREACHED */ } /* Already freed by callee */ goto senderr; } ip6 = mtod(m, struct ip6_hdr *); #endif /* PF */ error = nd6_output(ifp, origifp, m, dst, rt, NULL); if (error) { in6_ifstat_inc(ifp, ifs6_out_discard); ip6stat.ip6s_cantforward++; } else { ip6stat.ip6s_forward++; in6_ifstat_inc(ifp, ifs6_out_forward); if (type) ip6stat.ip6s_redirectsent++; else { if (mcopy) { goto freecopy; } } } #if PF senderr: #endif /* PF */ if (mcopy == NULL) { /* Release extra ref */ RT_REMREF(rt); return (NULL); } switch (error) { case 0: #if 1 if (type == ND_REDIRECT) { icmp6_redirect_output(mcopy, rt); /* Release extra ref */ RT_REMREF(rt); return (NULL); } #endif goto freecopy; case EMSGSIZE: /* xxx MTU is constant in PPP? */ goto freecopy; case ENOBUFS: /* Tell source to slow down like source quench in IP? */ goto freecopy; case ENETUNREACH: /* shouldn't happen, checked above */ case EHOSTUNREACH: case ENETDOWN: case EHOSTDOWN: default: type = ICMP6_DST_UNREACH; code = ICMP6_DST_UNREACH_ADDR; break; } icmp6_error(mcopy, type, code, 0); /* Release extra ref */ RT_REMREF(rt); return (NULL); freecopy: m_freem(mcopy); /* Release extra ref */ RT_REMREF(rt); return (NULL); }
/* * Ethernet output routine. * Encapsulate a packet of type family for the local net. * Use trailer local net encapsulation if enough data in first * packet leaves a multiple of 512 bytes of data in remainder. */ int ether_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst, struct route *ro) { short type; int error = 0, hdrcmplt = 0; u_char esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN]; struct llentry *lle = NULL; struct rtentry *rt0 = NULL; struct ether_header *eh; struct pf_mtag *t; int loop_copy = 1; int hlen; /* link layer header length */ if (ro != NULL) { if (!(m->m_flags & (M_BCAST | M_MCAST))) lle = ro->ro_lle; rt0 = ro->ro_rt; } #ifdef MAC error = mac_ifnet_check_transmit(ifp, m); if (error) senderr(error); #endif M_PROFILE(m); if (ifp->if_flags & IFF_MONITOR) senderr(ENETDOWN); if (!((ifp->if_flags & IFF_UP) && (ifp->if_drv_flags & IFF_DRV_RUNNING))) senderr(ENETDOWN); hlen = ETHER_HDR_LEN; switch (dst->sa_family) { #ifdef INET case AF_INET: if (lle != NULL && (lle->la_flags & LLE_VALID)) memcpy(edst, &lle->ll_addr.mac16, sizeof(edst)); else error = arpresolve(ifp, rt0, m, dst, edst, &lle); if (error) return (error == EWOULDBLOCK ? 0 : error); type = htons(ETHERTYPE_IP); break; case AF_ARP: { struct arphdr *ah; ah = mtod(m, struct arphdr *); ah->ar_hrd = htons(ARPHRD_ETHER); loop_copy = 0; /* if this is for us, don't do it */ switch(ntohs(ah->ar_op)) { case ARPOP_REVREQUEST: case ARPOP_REVREPLY: type = htons(ETHERTYPE_REVARP); break; case ARPOP_REQUEST: case ARPOP_REPLY: default: type = htons(ETHERTYPE_ARP); break; } if (m->m_flags & M_BCAST) bcopy(ifp->if_broadcastaddr, edst, ETHER_ADDR_LEN); else bcopy(ar_tha(ah), edst, ETHER_ADDR_LEN); } break; #endif #ifdef INET6 case AF_INET6: if (lle != NULL && (lle->la_flags & LLE_VALID)) memcpy(edst, &lle->ll_addr.mac16, sizeof(edst)); else error = nd6_storelladdr(ifp, m, dst, (u_char *)edst, &lle); if (error) return error; type = htons(ETHERTYPE_IPV6); break; #endif #ifdef IPX case AF_IPX: if (ef_outputp) { error = ef_outputp(ifp, &m, dst, &type, &hlen); if (error) goto bad; } else type = htons(ETHERTYPE_IPX); bcopy(&((const struct sockaddr_ipx *)dst)->sipx_addr.x_host, edst, sizeof (edst)); break; #endif #ifdef NETATALK case AF_APPLETALK: { struct at_ifaddr *aa; if ((aa = at_ifawithnet((const struct sockaddr_at *)dst)) == NULL) senderr(EHOSTUNREACH); /* XXX */ if (!aarpresolve(ifp, m, (const struct sockaddr_at *)dst, edst)) { ifa_free(&aa->aa_ifa); return (0); } /* * In the phase 2 case, need to prepend an mbuf for the llc header. */ if ( aa->aa_flags & AFA_PHASE2 ) { struct llc llc; ifa_free(&aa->aa_ifa); M_PREPEND(m, LLC_SNAPFRAMELEN, M_NOWAIT); if (m == NULL) senderr(ENOBUFS); llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP; llc.llc_control = LLC_UI; bcopy(at_org_code, llc.llc_snap_org_code, sizeof(at_org_code)); llc.llc_snap_ether_type = htons( ETHERTYPE_AT ); bcopy(&llc, mtod(m, caddr_t), LLC_SNAPFRAMELEN); type = htons(m->m_pkthdr.len); hlen = LLC_SNAPFRAMELEN + ETHER_HDR_LEN; } else { ifa_free(&aa->aa_ifa); type = htons(ETHERTYPE_AT); } break; } #endif /* NETATALK */ case pseudo_AF_HDRCMPLT: { const struct ether_header *eh; hdrcmplt = 1; eh = (const struct ether_header *)dst->sa_data; (void)memcpy(esrc, eh->ether_shost, sizeof (esrc)); /* FALLTHROUGH */ case AF_UNSPEC: loop_copy = 0; /* if this is for us, don't do it */ eh = (const struct ether_header *)dst->sa_data; (void)memcpy(edst, eh->ether_dhost, sizeof (edst)); type = eh->ether_type; break; } #ifdef MPLS case AF_MPLS: if (lle != NULL && (lle->la_flags & LLE_VALID)) bcopy(&lle->ll_addr.mac16 , edst, sizeof(edst)); else error = mpls_arpresolve(ifp, rt0, m, dst, edst, &lle); if (error) return (error == EWOULDBLOCK ? 0 : error); if (m->m_flags & (M_BCAST | M_MCAST)) type = htons(ETHERTYPE_MPLS_MCAST); else type = htons(ETHERTYPE_MPLS); break; #endif /* MPLS */ default: if_printf(ifp, "can't handle af%d\n", dst->sa_family); senderr(EAFNOSUPPORT); } if (lle != NULL && (lle->la_flags & LLE_IFADDR)) { update_mbuf_csumflags(m, m); return (if_simloop(ifp, m, dst->sa_family, 0)); } /* * Add local net header. If no space in first mbuf, * allocate another. */ M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT); if (m == NULL) senderr(ENOBUFS); eh = mtod(m, struct ether_header *); (void)memcpy(&eh->ether_type, &type, sizeof(eh->ether_type)); (void)memcpy(eh->ether_dhost, edst, sizeof (edst)); if (hdrcmplt) (void)memcpy(eh->ether_shost, esrc, sizeof(eh->ether_shost)); else (void)memcpy(eh->ether_shost, IF_LLADDR(ifp), sizeof(eh->ether_shost)); /* * If a simplex interface, and the packet is being sent to our * Ethernet address or a broadcast address, loopback a copy. * XXX To make a simplex device behave exactly like a duplex * device, we should copy in the case of sending to our own * ethernet address (thus letting the original actually appear * on the wire). However, we don't do that here for security * reasons and compatibility with the original behavior. */ if ((ifp->if_flags & IFF_SIMPLEX) && loop_copy && ((t = pf_find_mtag(m)) == NULL || !t->routed)) { if (m->m_flags & M_BCAST) { struct mbuf *n; /* * Because if_simloop() modifies the packet, we need a * writable copy through m_dup() instead of a readonly * one as m_copy[m] would give us. The alternative would * be to modify if_simloop() to handle the readonly mbuf, * but performancewise it is mostly equivalent (trading * extra data copying vs. extra locking). * * XXX This is a local workaround. A number of less * often used kernel parts suffer from the same bug. * See PR kern/105943 for a proposed general solution. */ if ((n = m_dup(m, M_NOWAIT)) != NULL) { update_mbuf_csumflags(m, n); (void)if_simloop(ifp, n, dst->sa_family, hlen); } else ifp->if_iqdrops++; } else if (bcmp(eh->ether_dhost, eh->ether_shost, ETHER_ADDR_LEN) == 0) { update_mbuf_csumflags(m, m); (void) if_simloop(ifp, m, dst->sa_family, hlen); return (0); /* XXX */ } } /* * Bridges require special output handling. */ if (ifp->if_bridge) { BRIDGE_OUTPUT(ifp, m, error); return (error); } #if defined(INET) || defined(INET6) if (ifp->if_carp && (error = (*carp_output_p)(ifp, m, dst))) goto bad; #endif /* Handle ng_ether(4) processing, if any */ if (IFP2AC(ifp)->ac_netgraph != NULL) { KASSERT(ng_ether_output_p != NULL, ("ng_ether_output_p is NULL")); if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) { bad: if (m != NULL) m_freem(m); return (error); } if (m == NULL) return (0); } /* Continue with link-layer output */ return ether_output_frame(ifp, m); }
/* * Ethernet output routine. * Encapsulate a packet of type family for the local net. * Use trailer local net encapsulation if enough data in first * packet leaves a multiple of 512 bytes of data in remainder. */ int ether_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *dst, struct route *ro) { short type; int error = 0, hdrcmplt = 0; u_char edst[ETHER_ADDR_LEN]; struct llentry *lle = NULL; struct rtentry *rt0 = NULL; struct ether_header *eh; struct pf_mtag *t; int loop_copy = 1; int hlen; /* link layer header length */ int is_gw = 0; uint32_t pflags = 0; if (ro != NULL) { if (!(m->m_flags & (M_BCAST | M_MCAST))) { lle = ro->ro_lle; if (lle != NULL) pflags = lle->la_flags; } rt0 = ro->ro_rt; if (rt0 != NULL && (rt0->rt_flags & RTF_GATEWAY) != 0) is_gw = 1; } #ifdef MAC error = mac_ifnet_check_transmit(ifp, m); if (error) senderr(error); #endif M_PROFILE(m); if (ifp->if_flags & IFF_MONITOR) senderr(ENETDOWN); if (!((ifp->if_flags & IFF_UP) && (ifp->if_drv_flags & IFF_DRV_RUNNING))) senderr(ENETDOWN); hlen = ETHER_HDR_LEN; switch (dst->sa_family) { #ifdef INET case AF_INET: if (lle != NULL && (pflags & LLE_VALID) != 0) memcpy(edst, &lle->ll_addr.mac16, sizeof(edst)); else error = arpresolve(ifp, is_gw, m, dst, edst, &pflags); if (error) return (error == EWOULDBLOCK ? 0 : error); type = htons(ETHERTYPE_IP); break; case AF_ARP: { struct arphdr *ah; ah = mtod(m, struct arphdr *); ah->ar_hrd = htons(ARPHRD_ETHER); loop_copy = 0; /* if this is for us, don't do it */ switch(ntohs(ah->ar_op)) { case ARPOP_REVREQUEST: case ARPOP_REVREPLY: type = htons(ETHERTYPE_REVARP); break; case ARPOP_REQUEST: case ARPOP_REPLY: default: type = htons(ETHERTYPE_ARP); break; } if (m->m_flags & M_BCAST) bcopy(ifp->if_broadcastaddr, edst, ETHER_ADDR_LEN); else bcopy(ar_tha(ah), edst, ETHER_ADDR_LEN); } break; #endif #ifdef INET6 case AF_INET6: if (lle != NULL && (pflags & LLE_VALID)) memcpy(edst, &lle->ll_addr.mac16, sizeof(edst)); else error = nd6_resolve(ifp, is_gw, m, dst, (u_char *)edst, &pflags); if (error) return (error == EWOULDBLOCK ? 0 : error); type = htons(ETHERTYPE_IPV6); break; #endif case pseudo_AF_HDRCMPLT: { const struct ether_header *eh; hdrcmplt = 1; /* FALLTHROUGH */ case AF_UNSPEC: loop_copy = 0; /* if this is for us, don't do it */ eh = (const struct ether_header *)dst->sa_data; (void)memcpy(edst, eh->ether_dhost, sizeof (edst)); type = eh->ether_type; break; } default: if_printf(ifp, "can't handle af%d\n", dst->sa_family); senderr(EAFNOSUPPORT); } if ((pflags & LLE_IFADDR) != 0) { update_mbuf_csumflags(m, m); return (if_simloop(ifp, m, dst->sa_family, 0)); } /* * Add local net header. If no space in first mbuf, * allocate another. */ M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT); if (m == NULL) senderr(ENOBUFS); eh = mtod(m, struct ether_header *); if (hdrcmplt == 0) { memcpy(&eh->ether_type, &type, sizeof(eh->ether_type)); memcpy(eh->ether_dhost, edst, sizeof (edst)); memcpy(eh->ether_shost, IF_LLADDR(ifp),sizeof(eh->ether_shost)); } /* * If a simplex interface, and the packet is being sent to our * Ethernet address or a broadcast address, loopback a copy. * XXX To make a simplex device behave exactly like a duplex * device, we should copy in the case of sending to our own * ethernet address (thus letting the original actually appear * on the wire). However, we don't do that here for security * reasons and compatibility with the original behavior. */ if ((ifp->if_flags & IFF_SIMPLEX) && loop_copy && ((t = pf_find_mtag(m)) == NULL || !t->routed)) { if (m->m_flags & M_BCAST) { struct mbuf *n; /* * Because if_simloop() modifies the packet, we need a * writable copy through m_dup() instead of a readonly * one as m_copy[m] would give us. The alternative would * be to modify if_simloop() to handle the readonly mbuf, * but performancewise it is mostly equivalent (trading * extra data copying vs. extra locking). * * XXX This is a local workaround. A number of less * often used kernel parts suffer from the same bug. * See PR kern/105943 for a proposed general solution. */ if ((n = m_dup(m, M_NOWAIT)) != NULL) { update_mbuf_csumflags(m, n); (void)if_simloop(ifp, n, dst->sa_family, hlen); } else if_inc_counter(ifp, IFCOUNTER_IQDROPS, 1); } else if (bcmp(eh->ether_dhost, eh->ether_shost, ETHER_ADDR_LEN) == 0) { update_mbuf_csumflags(m, m); (void) if_simloop(ifp, m, dst->sa_family, hlen); return (0); /* XXX */ } } /* * Bridges require special output handling. */ if (ifp->if_bridge) { BRIDGE_OUTPUT(ifp, m, error); return (error); } #if defined(INET) || defined(INET6) if (ifp->if_carp && (error = (*carp_output_p)(ifp, m, dst))) goto bad; #endif /* Handle ng_ether(4) processing, if any */ if (ifp->if_l2com != NULL) { KASSERT(ng_ether_output_p != NULL, ("ng_ether_output_p is NULL")); if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) { bad: if (m != NULL) m_freem(m); return (error); } if (m == NULL) return (0); } /* Continue with link-layer output */ return ether_output_frame(ifp, m); }